RetinaFace: Single-stage Dense Face Localisation in the Wild

By Jiankang Deng, Jia Guo, Jinke Yu, Irene Kotsia, Stefanos P. Zafeiriou. arXiv.org, 2019

747 citations recorded by the source corpus.

Though tremendous strides have been made in uncontrolled face detection, accurate and efficient face localisation in the wild remains an open challenge. This paper presents a robust single-stage face detector, named RetinaFace, which performs pixel-wise face localisation on various scales of faces by taking advantages of joint extra-supervised and self-supervised multi-task learning. Specifically, We make contributions in the following five aspects: (1) We manually annotate five facial landmarks on the WIDER FACE dataset and observe significant improvement in hard face detection with the assistance of this extra supervision signal. (2) We further add a self-supervised mesh decoder branch for predicting a pixel-wise 3D shape face information in parallel with the existing supervised branches. (3) On the WIDER FACE hard test set, RetinaFace outperforms the state of the art average precision (AP) by 1.1% (achieving AP equal to 91.4%). (4) On the IJB-C test set, RetinaFace enables state of the art methods (ArcFace) to improve their results in face verification (TAR=89.59% for FAR=1e-6). (5) By employing light-weight backbone networks, RetinaFace can run real-time on a single CPU core for a VGA-resolution image. Extra annotations and code have been made available at: this https URL.

Subfields: Computer Vision and Pattern Recognition, Signal Processing

View Paper: https://www.semanticscholar.org/paper/e8f20892d6df40ee7d42f36255a01fae67805233

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Computer Vision and Pattern Recognition

In Signal Processing

Citations in the corpus

Distract Your Attention: Multi-Head Cross Attention Network for Facial Expression Recognition (2021, methodology): “For the SFEW 2.0 dataset, the facial images are manually aligned using the RetinaFace [9] model.”
CN-Celeb: A Challenging Chinese Speaker Recognition Dataset (2019, methodology): “The RetinaFace algorithm was used to perform the detection and clipping [17].”
Detecting Deepfakes with Self-Blended Images (2022, methodology): “We adopt Dlib [34] and RetinaFace [18] to extract facial landmarks and bounding boxes from each video frame, respectively.”
CelebA-Spoof: Large-Scale Face Anti-Spoofing Dataset with Rich Annotations (2020, methodology): “The bounding box of faces are extracted by RetinaFace [8].”
General Facial Representation Learning in a Visual-Linguistic Manner (2021, methodology): “To filter out those non-face images, we adopt a face detector, RetinaFace [23], to identify the presence of a face in an image.”
Multi-attentional Deepfake Detection (2021, methodology): “For all real/fake video frames, we use a state-of-the-art face extractor RetinaFace[8] to detect faces and save the aligned facial images as inputs with a size of 380 × 380 .”
VarGFaceNet: An Efficient Variable Group Convolutional Neural Network for Lightweight Face Recognition (2019, methodology): “All face images in this dataset are aligned by five facial landmarks predicted from RetinaFace[5] then resized to 112 × 112.”
End-to-End Reconstruction-Classification Learning for Face Forgery Detection (2022, methodology): “For the FaceForensics++ [11], Celeb-DF [7] and DFDC [3] datasets, we use RetinaFace [2] to extract the face region from video sequences and adopt a conservative crop which enlarges the facial region by a factor of 1.”
Discriminability Distillation in Group Representation Learning (2020, methodology): “For data pre-processing, RetinaFace [11] is used to detect faces and their corresponding landmarks for all datasets.”
R2RNet: Low-light Image Enhancement via Real-low to Real-normal Network (2021, methodology): “Because there are no corresponding labels in the test set, we randomly select 100 images from the training set for evaluation, applying our R2RNet as a pre-processing step, followed by two state-of-art pre-trained face detection methods: RetinaFace [55] and DSFD [56].”
YOLO-FaceV2: A Scale and Occlusion Aware Face Detector (2022, background): “RetinaFace [45] applies independent context modules on five feature pyramid levels to increase the receptive field and enhance the rigid context modelling power.”
Synthetic Data for Deep Learning (2019, background): “• the labeling for some basic computer vision problems is especially complex: while pose estimation is doable, facial keypoint detection (a key element for facial recognition and image manipulation for faces) may require to specify several dozen landmarks on a human face, which becomes very hard for”
Sub-center ArcFace: Boosting Face Recognition by Large-Scale Noisy Web Faces (2020, methodology): “We build our face recognition system by RetinaFace (ResNet-50) [7] and sub-center ArcFace (ResNet-100), and accelerate the inference by the openVINO toolkit.”
Self-Supervised Learning of Audio-Visual Objects from Video (2020, methodology): “As a face detector baseline, we use the state-of-the-art RetinaFace [16] detector, with both the MobileNet and ResNet-50 backbones.”